Joint Constraint Aggregation Reduces Retrieval Skew in RAG Systems

Joint Constraint Aggregation (JCA) improves RAG accuracy by comparing candidates against question-derived requirements rather than relying on candidate-level support. This method mitigates errors where retrieved evidence frequently mentions a competing answer that satisfies only some question constraints.